Looks like it's not just for JS pages, e.g. https://global-search.toolforge.org/?q=WSexport results in

The server said: cURL error 6: Could not resolve host: cloudelastic1004.wikimedia.org (see https://curl.haxx.se/libcurl/c/libcurl-errors.html) for https://cloudelastic1004.wikimedia.org:8243/*_content,*_general,*_file,*:*_content,*:*_general,*:*_file/_search

cloudelastic1004.wikimedia.org

See T358046: decommission cloudelastic100[1-4].wikimedia.org

Okay, we're back up and running. I do not recall why we were querying cloudelastic1004.wikimedia.org directly instead of https://cloudelastic.wikimedia.org.

We're now querying the "chi" cluster at https://cloudelastic.wikimedia.org:8243. According to the docs, this includes "approximately the 200 largest wikis", but is missing some of the smaller wikis. I don't know if this means we have potentially incomplete results, compared to what we were getting before on cloudelastic1004.

Perhaps @EBernhardson could shed some light on how it all works. In order to be truly "global", should we be querying all three clusters and join the results?

I suspect at the time we initially setup global-search we didn't have the cloudelastic.wikimedia.org alias up and running yet, but now that that exists should certainly point at it instead of individual servers.

In terms of the other two search clusters, you are querying them through the cross-cluster search functionality. At query time the syntax *_content is searching indicies in the local cluster, and *:*_content is searching indices available over cross-cluster search.